Gemini Cuts Video Tokens 88% With Agentic Video Understanding
Google has added a new processing mode to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite that lets the model pick out only the parts of a video it actually needs.
Google's AI research organization that builds Gemini. AlphaGo and AlphaFold also came out of here. In August 2026, Google released Gemini 3.7 Flash, a coding-focused model, surpassing Anthropic's Claude on benchmarks. That same month, Gemini 3.7 Flash was updated again to a version that raised its intelligence score while cutting response time. Pixel Watch 5 will feature an offline-capable version of Gemini, with its price rising by $50. In August 2026, Demis Hassabis stepped down from Google DeepMind and Jeff Dean also left the company after 27 years, sending Alphabet's stock down 4%.
Current rank (1M)
3
Rank over the last 30 days
2026.08.18 – 2026.09.16 · High 3 · Low 5 · Now 3
Google has added a new processing mode to Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite that lets the model pick out only the parts of a video it actually needs.
GenAI.mil has added ChatGPT Mil and Grok for Government, but Anthropic's Claude is still missing.
A Labs prototype bundling paper tracking, code conversion, and hypothesis generation
In an interview video released by Google DeepMind, the robot Apollo said tying a trash bag was harder than balancing during sports
A VP of Research at Google DeepMind discussed how to handle uncertainty, from weather forecasting to robotics
Google Research sends congratulations as Dr. Croak is recognized as a pioneer in AI
An encryption method that keeps both model weights and test questions hidden from each other cuts the risk of benchmark manipulation
In an interview, Thibault lays out pricing and agent UX strategy
Claude has Chat and Cowork, ChatGPT has Chat and Work — this is turning into an industry-wide problem for AI apps
The foundation model trains only on continuous glucose monitoring data and predicts diabetes risk more accurately than existing models
Google DeepMind revealed a custom vocabulary screen, pointing toward better recognition accuracy for specific terms
Live and Live Experimental also launched alongside it, but the Gemini 3.5 Pro model teased back in June still hasn't shown up